一种用于水下图像的光谱重构方法:理论及应用

多光谱成像是一项非常有前景的图像高保真获取与再现技术，近年来在水下物体颜色还原的应用中也受到的极大的需求和关注。然而，不同于空气中的物体的成像过程，在水下成像过程中，当光通过水而进行传播，光被水体严重吸收和散射，导致图像变暗，在其光谱和颜色方面发生模糊和扭曲。文中讨论的是基于水下图像的水衰减系数的校准和其多光谱图像的光谱重构。首先在不同的距离处获取物体的图像，提出了基于不同距离的图像进行水体衰减系数的校准并恢复原始图像的技术；在此基础上，分析并导出满足系数校准和图像复原所需的在不同距离获取到的最少的原始图像个数。最后，通过比较复原的水下图像与空气中获取的彩色图像，实验结果证明:文中提出的技术能够对水下光谱图像的进行精确颜色复原，所有测试图像的平均相对残留误差仅为5.87％。

Method for spectral restoration of underwater images: theory and application

Underwater multispectral imaging is a promising technique for high-fidelity underwater color reproduction and mapping of kelp, sea grass, corals, etc. However, as light propagates through water, light is severely absorbed and scattered by water, causing image dim, hazy and distorted in its spectrum and color. In this paper, calibration of water attenuation coefficient based on underwater images and restoration of underwater multispectral images are discussed. Multispectral images of an underwater object are captured at different underwater distances. Technique has been proposed to calibrate the water attenuation coefficient based on underwater images of different distances and restore the raw images. Analysis was also conducted to search for the least number of distances for coefficient calibration and restoration. By comparing the restored underwater images with the images captured in air, its found that the technique proposed in this paper provides accurate restoration of underwater spectral images, with a relative residual error of 5.87% in average for all test images.